license: creativeml-openrail-m
I preferred using DPM++ 2M SDE Karras for my model, but it's not necessarily the best
sdxl usage is completely different from all previous models.
1/ the number of cfg is greatly affected, can replace clip skip. if you want to create surreal images, cfg 3-4 is suitable, cartoon or 3d character images should be clear, cfg should be 5-6, and real people photos can sometimes be up to 7-8 but there are It can be lower, but only if you need high precision and detail to get cfg up to 9-10 or higher, I haven't tried it yet.
2/ the number of steps also has a much greater influence on the results, the image only starts to achieve high quality when it is 50 steps or more. Right now I don't know where its limit is, it's like the higher the step count, the better the image. Of course it can be different when changing styles, changing models, or even changing samplers. I still can't try it all.
3/ the way to get prompts will be different from other sd models, but I don’t know how, just feel that the prompts are too cumbersome and not as good as before and instead of writing in tag style, You can try to choose a descriptive writing style like Midjourney
4/ That's all for now, I'll keep updating as new information or changes become available, anyway it's new technology for all of us.
